Multiplexing the TC port - Is it feasible?


 
T

Tony L-Iowa

TVWBB Fan
Curious if you all think this idea would be feasible as a "future upgrade".

The HM already contains an optional single TC port (port 0) with the added SMC amplifier circuitry. Would it be feasible instead of installing the thermocouple jack to instead break out the TC signals and perhaps an additional port expansion selector signal to an external add on box that contained a set of thermocouple jacks and a multiplexer chip to route the selected port signals back to the currently existing TC amp circuit? Then, use the Pi/ATMega to control the port switching and data collection process of going between multiple external TC ports? It may require something like a new HM serialized constrol/switching signal to the external demuxer/muxer to keep it down to minimal wires (e.g. power, gnd, serialized control signal, TC +/- feedback signals).

I could see the TC sensor sampling times would likely slow down based on the number of external ports expanded but can anyone else think of reasons why this may or may not be feasible without needing to change the HM circuitry and current board design? It sure seems like a potentially viable way to get expanded TC port support without spending a ton of extra money (i.e. some wires, external board with mux/demux chips and TC ports). I could see most of the work would likely be in software updates to do the switching/sampling logic and changes to control how many TC ports are external switched.

Thoughts?
 
I think one of the downsides to multiplexing the temperature probes is that it would drastically reduce the sample rate that the heatermeter currently uses, which is 512 samples per second, per probe.
 
It should be doable in a way that don't affect the sampling rate of currently existing ports (except for port0 when extended TC port activation is in effect). The rest of the TC extension ports could become ports 4 through X and be round robin sampled with the TC port0 port. Usually we're dealing with BBQ cooking on the order of hours here. I wouldn't think lowering the sample rates of "some of the ports" would do much harm if the user chose to activate such a feature. If the math/algorithms were tightly coupled to 512 sample rate then just fill the between data on the down-sampled ports with the previously sampled/refreshed value until the next time it actually gets newly measured data from the sensor. I haven't seen the code in this area so I'm not exactly sure what it's design looks like under the hood.
 
I actually have thought about this as well, some way of having an analog front end that can switch in and out lines and multiplex the thermocouple and the thermistor ports so they can be mixed and matched. My first experiment would be to see if the thermocouple reads properly while connected to an analog mux (which has its own resistance as well as may not be symmetric). I got the parts needed to make that work this week but I've had my hands full with the ZeroW stuff.

The downside is that a through hole 8 x 2 analog mux is going to be a pretty giant part so there is no easy way to fit it on the existing board without adding another 25% to the size. Also it means that everything has to be that way and the sampling core in the firmware will need to possibly only support that rather than being able to work both ways so it would break compatibility.

It is most likely possible, but what's $10.50 extra worth of thermocouple amps compared to the overall cost (and the connector alone is $2 each, which you need either way).

EDIT: And we actually do 255 samples times 4 probes, about 6 times every second. Nobody samples more than we do!
 
Last edited:
I was thinking you'd put the mux/demux in another external board/box and just run 5-6 wires out to it using current TC slot or making a new case hole for the new wires (small voltage signal wire to power the external mux/demux chips, gnd wire, serial control wires to switch the mux port selections, and return wires for the currently active ThermoCouple positive/negative). If done this way current users of HeaterMeter could do this upgrade without building a whole new unit and it comes available as a software upgrade with some tweaks to soldering up a few new wire connections. Also, are the TC type K mini connectors really necessary @ $2 each? I've seen some cheap (http://www.ebay.com/itm/like/381675830556?lpid=82&chn=ps&ul_noapp=true) just screw down type wire connectors.

Here's a discussion on TC muxed switching:
http://electronics.stackexchange.com/questions/182646/understanding-slow-response-of-multiplexed-thermocouples

Here's kind of an off the shelf card doing this already:
https://oceancontrols.com.au/KTA-259.html
 
The idea of the 'external board box' would have some issues. The TC +/- cables can only be extended with the same TC cable-type material. So you'd have to run data lines and a thermo couple cable to the box that has the mux/demux. IMO the best is to use SPI to a MAX TC/mux external board therfore you are just using data lines to the box.

Also, from others and looking at what others are using. Which analog mux is important apparently...some just didn't work. TC really needs/wants a 'solid-same-material' connection to the IC. One reseller of a tc multiplex board uses this chip: ADG608 Multiplexer IC which I assume works for them.

I found this chip that seems like it may be a good solution for upt to 4 TCs .... Bryan have you seen the spec sheet on it? Can it be used to support both ADC and TC at the same time as it states config per channel?
 
I've not really looked too much into it just because I wouldn't ever use it. Thermistor probes work just fine for food probes and that's what I have. All my thermocouples are blunt-tipped, alligator-clipped, or bare wires so I would have to get all new food probes too. Others have asked about it which is why it is on my idle experimentation list, and I know it can be done but the question is if it is cost effective and beneficial enough to include.

What I bought was a TS3A5017PWR which is cheap and switches both lines together, but like I said I haven't tried it yet I just have selected the part. That ADG608 is pretty expensive if all you need is an analog mux. That AD7124-8BCPZ chip looks alright in that is has the integrated drivers, enough inputs, and high enough gain that we can get to where we are now (125x). You'd need to implement your own cold junction compensation though in software, which is why I would think a mux and a dedicated TC amp with intregrated compensation might be better.

@Tony: I've not seen any board mount thermocouple connectors anywhere apart from the PCC-SMP-K we use actually. I hate having a part where there is just one person who makes it and only one person that sells it for a non-insane price ($2 instead of $5/ea), but I've never seen another. Thermoworks uses panel mount jacks in their devices and have wires run to the amplifier. Cheap devices have these prongs that solder directly to the board which would work but I've never seen them for sale.
 
Yah, it's always been puzzling why Newark and Omega where the only two places you could get the TC connectors. I just did a fresh search and to my surprise Mouser seems to be carrying them now, at the Mouser price of coarse and apparently small quantities.
http://www.mouser.com/ProductDetail...S1kDzRVYBg==&gclid=CJiAi7_ChtMCFZaEaQod3pEPAg
Sparkfun also seems to be carrying them as well as a few other companies that are now popping up, unfortunately they are all more expensive than Newark/Omega.
Hopefully sales at Mouser will jump and they will stock more and lower the price, which will at least eliminate that extra shipping cost for the single TC connector if you are already ordering from either of these companies.
 
You must log in or register to reply here.

 

Back
Top